Work

Development and internal validation of ultrasound microvascular flow imaging-based radiomics models for predicting BRAF V600E mutation status in thyroid nodules

Year: 2026

Type: article

Abstract: Background/objectives To develop and internally validate a microvascular flow imaging (MFI)-based radiomics model for the preoperative prediction of BRAF V600E mutation status in thyroid nodules (TNs). Methods Our research encompassed 170 patients with TNs who experienced traditional ultrasound (US), MFI, and BRAF V600E testing. Patients were randomly divided into a training cohort (n = 136) and an internal validation cohort (n = 34). Radiomics characteristics were obtained from MFI images following manual segmentation, and reproducibility was assessed using intraclass correlation coefficients. Significant clinical and US variables were incorporated into radiomics characteristics. Feature selection (FS) was undertaken via LASSO regression, followed by machine-learning model construction. A radiomics-only model (ROM) was also developed. Model performance was evaluated using the area under the receiver operating characteristic curve (AUC), accuracy, sensitivity, and specificity. Calibration was assessed using calibration curves, and potential clinical utility was explored using decision curve analysis (DCA). Results Significant differences were observed between the BRAF V600E mutation-positive and mutation-negative groups in Bethesda category, morphology, age, echogenicity, and vascularity (all P < 0.05). A total of 18 features were selected, including 15 MFI-derived radiomics features, two conventional ultrasound features, and one cytopathological feature (Bethesda category III). In the internal validation cohort, the combined model (CM) achieved an AUC of 0.854 (95% CI: 0.728–0.979), compared with 0.764 (95% CI: 0.596–0.933) for the ROM. The CM showed a numerically higher AUC, although statistical superiority was not established. Conclusions The CM integrating MFI-based radiomics with clinical and conventional ultrasound characteristics showed promising performance in predicting BRAF V600E mutation status. It may serve as an adjunctive preoperative prediction tool to support individualized clinical decision-making in patients with thyroid nodules. However, because the CM incorporates Bethesda cytological classification, it should not be considered a fully non-invasive model. Further independent validation is warranted.

Source: Journal of Radiation Research and Applied Sciences

DOI: 10.1016/j.jrras.2026.102793

Authors: Jing Wang, Fei Chen, Yu Dou

Institutions: First Hospital of Lanzhou University, Lanzhou University

Language: English

Topic: Radiomics and Machine Learning in Medical Imaging

Subfield: Radiology, Nuclear Medicine and Imaging

Field: Medicine

Domain: Health Sciences

SDG: Good health and well-being

Open Access status: gold

Estimated APC paid: $1,400